In task scheduling, the job-shop scheduling problem is notorious for being a combinatorial optimization problem; it is considered among the largest class of NP-hard problems. In this paper, a parallel implementation of hybrid cellular genetic algorithm is proposed in order to reach the best solutions at a minimum execution time. To avoid additional computation time and for real-time control, the fitness evaluation and genetic operations are entirely executed on a graphic processing unit in parallel; moreover, the chosen genetic representation, as well as the crossover, will always give a feasible solution. In this paper, a two-level scheme is proposed; the first and fastest uses several subpopulations in the same block, and the best solutions migrate between subpopulations. To achieve the optimal performance of the device and to reshape a more complex problem, a projection of the first on different blocks will make the second level. The proposed solution leads to speedups 18 times higher when compared to the best-performing algorithms.
Hybridization of Genetic Algorithm with Parallel Implementation of Simulated Annealing for Job Shop Scheduling
